No announcement yet.

Koloratur - a Danish-American active speaker

  • Filter
  • Time
  • Show
Clear All
new posts

  • Koloratur - a Danish-American active speaker

    Click image for larger version

Name:	image_65708.jpg
Views:	3
Size:	414.5 KB
ID:	1325349

    This story started with me poking around the Parts Express website 6 months ago - a budding speaker builder with only a couple of builds under my belt: A pair of Overnight Sensations in Zebrawood built recently, and a set of KEF speakers built ages ago during my junior high-school days growing up in Denmark. When I stumbled upon a factory buyout of the Vifa D19TD-27-08 tweeter made in the original factory in Videbęk, Denmark, about 70 miles from my little hometown, I knew I had to design a speaker around it.

    I'm a Danish-American (amateur) choral singer with an Italian-America wife who is a coloratura soprano, so reproduction of voices and accurate imaging were obvious top priorities. The project name "Koloratur" - the Danish word for coloratura - also came naturally.

    The little Vifa isn't happy going much below 3 kHz, and I didn't want to embark on a three-way for my first design, so I decided to go with the new Dayton Audio RS-150P-4A paper cone woofer that has a smooth frequency response and is small enough to have reasonable dispersion above 3 kHz. So like my marriage, this project is proudly combining parts that originate in Jylland (the western peninsula of Denmark) and the American Midwest.

    After running simulations in WinISD, I decided to sacrifice some max SPL in return for bass extension and chose a ported design tuned at 48Hz using the 0.56 cubic feet pre-made cabinet from PE. As you can see from the picture, I mounted the drivers as closely as possible, asymmetrically on the baffle to (hopefully) smooth out any diffraction bumps. I applied some extra (red oak) bracing and bitumen sheets to deaden the baffle and cabinet, and lined it with sonic barrier.

    Finally, to make it easy to tinker, tune, and learn, I chose to go the active route with a miniDSP 2x4 HD active crossover, and also kept things modular and flexible by not putting any electronics in the speaker cabinets (4-pole speakON connectors are awesome). I ended up crossing them at 3.5 kHz with 48 dB/octave slopes and applied bafflestep compensation and various corrections to straighten the response of the drivers. I also added a 30 Hz 48 dB/octave high-pass filter on the woofers to limit excursion from low-frequency material. I know I'll probably keep tinkering with the filters settings forever, but I have attached a curve measured in my garage at ~1m with the umik1 and REW reflecting the current state of affairs. It's reasonably flat from 40 Hz to 20kHz. I have the opportunity to measure them in an anechoic chamber soon, so hopefully I can get a clearer picture then.

    After some basic "dialing in", I'm quite impressed with the result (I'm obviously biased). The imaging is really pinpoint accurate and has great depth. Voices and piano sound beautifully clean and detailed. In rock & pop I discover quite a few new details and instruments are clean and nicely separated.

    The next step is choosing a finish and make them look as pretty as they sound.

    P.S. If you look carefully, you'll notice that the (excellent) beer consumed during the making of these speakers is from Grand Rapids, MI.
    Last edited by rmlarsen; 03-24-2017, 02:12 PM.

  • #2
    Originally posted by rmlarsen View Post
    This story ...

    After running simulations in WinISD, I decided to sacrifice some max SPL in return for a bit of bass extension ...
    Great story ...

    Perhaps the miniDSP can limit power in the freq. range that breaches xmax (don't know). At lower levels there isn't any effect At higher power the miniDSP insures the driver doesn't distort.


    • #3
      Millstonemike Good suggestion. I'll look into that. I did add a sharp 48dB/octave high-pass filter at 30 Hz on the woofers to limit xmax, but perhaps it's possible to do something dynamic with the miniDSP like you suggest.
      Last edited by rmlarsen; 03-01-2017, 12:16 PM.


      • #4
        Great story its nice when something is more than just a object.
        I currently live in the Grand Rapids area and it is a great place to live, With LOTS o good beer.


        • #5
          Very nice project and implementation!


          • #6
            Nice project. Did you run out of EQ settings for the woofer? It looks like it could use a little more help around 420Hz if you have any left. Like you said, you may find yourself tinkering with the DSP forever. It's easy and loads of fun. Careful, Active is considered the "Dark Side" of speaker building.
            My "No-Name" CC Speaker
            Kerry's "Silverbacks"
            Ben's Synchaeta's for Mom
            The Archers
            Rick's "db" Desktop CBT Arrays
            The Gandalf's


            • #7
              Thanks, guys.

              Kevin K. I spent more time massaging the tweeter response. I have tons of EQ filters left to throw at the woofer. I wasn't sure if the 420 HZ dip was a floor bounce or not, but will definitely take another look . I'm sure there are plenty of improvements to be had yet.

              BTW: I discovered that there appears to be a bug in the miniDSP DAC code. In the attached plot of the measured tweeter response, the red curve is measured feeding the signal from REW directly to miniDSP USB DAC input, while the green curve is using a FiiO K-1 USB DAC fed to the analog inputs of the miniDSP. That deep knotch at ~19 kHz looks bizarre, eh? At first I thought I'd broken my tweeter. I'm filing a bug report with miniDSP. It could also be an OS driver issue, I suppose, but I'd definitely be bummed if I paid $100 extra for a broken DAC in the HD model.


              • #8
                DSP is fun and powerful. Your system will have a power response dip in the 1500-2500hz range, in practice it's not the worst thing in the world, but as you move forward, one of the key goals (and one of the advantages of three-ways and directivity-matched waveguide systems) a flatter power response is a good goal to keep in mind.


                • #9
                  badman I'm new to this, so thanks for the advice. By power response do you mean total radiated power (integrated over 4 \pi steradians?) as opposed to the on-axis frequency response?


                  • #10
                    badman Found an explanation of the term confirming my guess: Thank you, Google! :-)


                    • #11
                      Paradigm (and some other NRC style speakers) used to run their tweeter XOs very low to mitigate this issue with large midwoofers. Horrible decision, as their Monitor 9 which used 8"s and a titanium dome always sounded strained because the tweeter was working so darned hard.


                      • #12
                        badman Yes, I definitely hear some strain even when crossing the Vifa at 3kHz instead of 3.5. I need to run some distortion and off-axis response measurements to map this out and make sure I'm not just kidding myself.


                        • #13
                          I'm a fan of using a small mid (3-4") any time I use a dome tweet, so you can cross somewhat higher without as severe a power response dip, and either use stereo subs crossed high (120hz) or a large midbass below. 300-500hz L/M crossover, 3-4khz M/H tends to give you a good balance of output, distortion profiles, and crossovers pushed away from the most sensitive midband. It will usually still have a power response dip below the tweeter XO but it won't be too severe.

                          If you're okay with laserbeam treble, there's a school of thought for using a 3-4" midtweet crossed in the 500hz range, and simply tolerate the top 2 octaves beaming. Many small "full" rangers (fostex) have rising responses, so you can skip the toe-in, and still get a good direct/reflected balance and on-axis performance at the sweet spot- but it's a 1 seat game.

                          Many have moved/are moving to constant directivity systems to allow a nice large sweetspot, crossing the axes in front of the "main" listening postition means that off-axis dropoff aligns to axial distance and makes for a balanced tonal/imaging perfomance across a couch-sized listening area. This is the approach I use in the "big rig", it was popularized in the studio monitor sector and geddes, parham, and others have helped transition it to home.


                          • #14
                            badman Funny you should mention "the other school of thought", since I just finished building an LXmini clone. I do find the small sweetspot to be a nuisance, and definitely prefer the Koloratur for pinpoint imaging, depth and clean upper range. I followed the LXMini clone build detailed here: It probably goes too cheap on the full range driver - it doesn't sound that clean even after being EQ'd to a very flat response, and the imaging is nothing like the real LXmini that I heard recently. I might spring for the SEAS full range driver or perhaps try one of the Tang Band ones.


                            • #15
                              Great first project! Active is definitely the way to go for your first project.

                              I'd recommend not using a 48dB/oct crossover slope because of the large phase shifts introduced by such steep filters. 24dB/oct is more than enough.

                              I'd leave the 420Hz dip alone. Looks like you're doing an in room measurement without gating. That's probably a room reflection induced dip.

                              Try using a low shelf filter to bring the bass up a bit. Move the high pass filter to 40Hz to better protect the speaker.

                              Have fun playing around!